Neodymium isotopic compositions (εNd) have been largely used for the last fifty years as a tracer of past ocean circulation, and more intensively during the last decade to investigate ocean circulation during the Cretaceous period. Despite a growing set of data, circulation patterns still remain unclear during this period. In particular, the identification of the deep-water masses and their spatial extension within the different oceanic basins are poorly constrained. In this study we present new deep-water εNd data inferred from the Nd isotope composition of fish remains and Fe–Mn oxyhydroxide coatings on foraminifera tests, along with new εNd data of residual (partly detrital) fraction recovered from DSDP Sites 152 (Nicaraguan Rise), 258 (Naturaliste Plateau), 323 (Bellinghausen Abyssal Plain), and ODP Sites 690 (Maud Rise) and 700 (East Georgia Basin, South Atlantic). The presence of abundant authigenic minerals in the sediments at Sites 152 and 690 detected by XRD analyses may explain both middle rare earth element enrichments in the spectra of the residual fraction and the evolution of residual fraction εNd that mirror that of the bottom waters at the two sites. The results point towards a close correspondence between the bottom water εNd values of Sites 258 and 700 from the late Turonian to the Santonian. Since the deep-water Nd isotope values at these two sites are also similar to those at other proto-Indian sites, we propose the existence of a common intermediate to deep-water mass as early as the mid-Cretaceous. The water mass would have extended from the central part of the South Atlantic to the eastern part of proto-Indian ocean sites, beyond the Kerguelen Plateau. Furthermore, data from south and north of the Rio Grande Rise–Walvis Ridge complex (Sites 700 and 530) are indistinguishable from the Turonian to Campanian, suggesting a common water mass since the Turonian at least. This view is supported by a reconstruction of the Rio Grande Rise–Walvis Ridge complex during the Turonian, highlighting the likely existence of a deep breach between the Rio Grande Rise and the proto-Walvis Ridge at that time. Thus deep-water circulation may have been possible between the different austral basins as early as the Turonian, despite the presence of potential oceanic barriers. Comparison of new seawater and residue εNd data on Nicaraguan Rise suggests a westward circulation of intermediate waters through the Caribbean Seaway during the Maastrichtian and Palaeocene from the North Atlantic to the Pacific. This westward circulation reduced the Pacific water influence in the Atlantic, and was likely responsible for more uniform, less radiogenic εNd values in the North Atlantic after 80Ma. Additionally, our data document an increasing trend observed in several oceanic basins during the Maastrichtian and the Palaeocene, which is more pronounced in the North Pacific. Although the origin of this increase still remains unclear, it might be explained by an increase in the contribution of radiogenic material to upper ocean waters in the northern Pacific. By sinking to depth, these waters may have redistributed to some extent more radiogenic signatures to other ocean basins through deep-water exchanges.